In response to stimulation with proinflammatory cytokines, the deubiquitinase A20 inducibly interacts with the regulatory molecules TAX1BP1, Itch and RNF11 to form the A20 ubiquitin-editing complex. However, the molecular signal that coordinates the assembly of this complex has remained elusive. Here we demonstrate that TAX1BP1 was inducibly phosphorylated on Ser593 and Ser624 in response to proinflammatory stimuli. The kinase IKKα, but not IKKβ, was required for phosphorylation of TAX1BP1 and directly phosphorylated TAX1BP1 in response to stimulation with tumor necrosis factor (TNF) or interleukin 1 (IL-1). TAX1BP1 phosphorylation was pivotal for cytokine-dependent interactions among TAX1BP1, A20, Itch and RNF11 and downregulation of signaling by the transcription factor NF-κB. IKKα therefore serves a key role in the negative feedback of NF-κB canonical signaling by orchestrating assembly of the A20 ubiquitin-editing complex to limit inflammatory gene activation.
Inflammation is critical for the host response to injury or infection with microbes. Cells of the innate immune system, such as macrophages, produce proinflammatory cytokines such as tumor necrosis factor (TNF) or interleukin-1β (IL-1β), which have important roles in clearing pathogens and initiating an adaptive immune response. Inflammatory responses are tightly regulated to prevent chronic inflammation, which has been linked to autoimmune diseases and many cancers1.
Transcription factors of the NF-κB (Rel) family act as master regulators of genes encoding molecules that control innate and adaptive immune responses2. NF-κB signaling consists of two distinct pathways: the canonical pathway and the noncanonical pathway. In the canonical NF-κB pathway, heterodimeric NF-κB proteins such as RelA (p65) and p50 are sequestered in the cytoplasm as inactive subunits by inhibitory IκB proteins. A wide variety of stimuli, such as TNF and IL-1, as well as microbial components, including lipopolysaccharide (LPS), trigger signaling pathways that culminate at the level of IκB kinase (IKK)3. IKK consists of the catalytic kinase subunits IKKα and IKKβ and the regulatory subunit IKKγ (NEMO)4,5. IKKβ and IKKγ serve important roles in the activation of the canonical pathway via phosphorylation of the NF-κB inhibitor IκBα, which leads to its ubiquitination and degradation by the 26S proteasome complex6. The degradation of IκBα allows NF-κB to enter the nucleus, where it regulates the expression of large numbers of target genes encoding molecules that regulate inflammation and survival.
In the noncanonical NF-κB pathway, various ligands of the TNF superfamily, such as BAFF and CD40L, trigger processing of the precursor protein p100 (NF-κB2) to yield the NF-κB subunit p52 (refs. 7,8). RelB and p52 regulate genes encoding molecules involved in B cell survival and lymphoid organogenesis. IKKα is activated in the noncanonical pathway by the kinase NIK (NF-κB-inducing kinase)9,10.
Activation of canonical NF-κB signaling is under tight control by many mechanisms, all of which ensure transient NF-κB signaling in response to a specific stimulus. The gene encoding IκBα is a target of NF-κB, and IκBα functions in a negative feedback loop to terminate NF-κB signaling11. In addition, the deubiquitinases CYLD and A20 target important signaling intermediates upstream of IκBα to terminate NF-κB activation12,13. CYLD and A20 cleave Lys63 (K63)-linked polyubiquitin chains on overlapping substrates, including the ubiquitin ligases TRAF2 and TRAF6 and the adaptor RIP1 (refs. 14,15,16). However, given the lack of redundancy between CYLD and A20, each of these deubiquitinases probably has a highly specific role in vivo in the regulation of inflammatory, apoptotic and developmental signaling pathways17,18,19,20.
Polymorphisms in the gene encoding A20 have been linked to many inflammatory autoimmune diseases in humans21. Furthermore, A20-deficient mice succumb to uncontrolled and spontaneous multiorgan inflammation as a result of unchecked NF-κB signaling22. Studies indicate that A20 also functions as a tumor-suppressor gene in several subtypes of B cell lymphomas23,24. Therefore, understanding the mechanisms of A20-mediated downregulation of NF-κB is of critical importance. A20 has been shown to function as a unique dual-function deubiquitinase and E3 ligase in the TNF receptor pathway25. A20 first cleaves K63-linked polyubiquitin chains on RIP1 via the OTU deubiquitinase domain, followed by zinc finger 4–mediated catalysis of K48-linked polyubiquitin chains, to trigger RIP1 degradation25. Another important mechanism used by A20 to downregulate NF-κB signaling is the disruption of interactions among the E2 conjugating enzymes Ubc13 and Ubc5hc and E3 ligases TRAF6, TRAF2, cIAP1 and cIAP2 (refs. 26,27).
It is becoming increasingly clear that A20 requires essential contributions from many other proteins, including the regulatory molecules TAX1BP1, Itch and RNF11 (refs. 26,28). In response to TNF or IL-1 or bacterial LPS, A20, TAX1BP1, Itch and RNF11 inducibly interact to form the A20 ubiquitin-editing complex28. Cells lacking either TAX1BP1, Itch or RNF11 are a phenocopy of A20-deficient cells in their persistent activation of NF-κB signaling29. Although the specific contributions of each of these proteins are unclear, TAX1BP1 probably functions as a ubiquitin-binding adaptor molecule for A20 (ref. 30). However, it is unknown what regulates the inducible interactions among A20, TAX1BP1, Itch and RNF11, a key event required for the termination of NF-κB signaling.
Human T cell leukemia virus type I (HTLV-I) is an oncogenic retrovirus etiologically linked to adult T cell leukemia-lymphoma31. HTLV-I encodes the oncogenic regulatory protein Tax, which persistently activates NF-κB to transform T cells and maintain the survival of virus-transformed cells32,33. Tax activates NF-κB by associating with IKKγ and triggering persistent activation of the IKK subunits34. Furthermore, Tax bypasses the negative feedback inhibition of NF-κB by interacting with TAX1BP1 and disabling the function of the A20 ubiquitin-editing complex35. The exact mechanism by which Tax inactivates A20 has remained elusive.
In this report we describe IKKα as a regulator and kinase for TAX1BP1. We found that TAX1BP1 was inducibly phosphorylated at Ser593 and Ser624 in response to stimulation with TNF and IL-1. Phosphorylation of TAX1BP1 was critical for the inducible interactions among proteins of the A20 ubiquitin-editing complex and the termination of TNF- and IL-1-dependent NF-κB signaling.
TNF and IL-1 induce phosphorylation of TAX1BP1 by IKKα
TAX1BP1 has an essential role in the termination of NF-κB signaling by regulating the recruitment and function of A20 (ref. 29). However, it is unclear if TAX1BP1 undergoes any type of post-translational modifications that may regulate its function. In response to stimulation with TNF or LPS, a form of TAX1BP1 that migrated slower was apparent in mouse embryonic fibroblasts (MEFs; Fig. 1a). Treatment of protein lysates with calf intestinal alkaline phosphatase converted the more slowly migrating form of TAX1BP1 to a form that migrated faster (Fig. 1b), which indicated that the shift in mobility was indeed caused by phosphorylation. We also observed phosphorylation of endogenous TAX1BP1 in MEFs labeled with 32P orthophosphate and treated with TNF, but not in unstimulated cells (Fig. 1c). We also observed by immunoblot analysis a change in TAX1BP1 mobility when TAX1BP1 was expressed together with either IKKα or IKKβ, but not when it was expressed together with IKKγ (Fig. 1d). However, phosphorylation of TAX1BP1 was intact in response to TNF and IL-1 in MEFs deficient in IKKβ (Ikbkb−/−) or IKKγ (Ikbkg−/−), but not in MEFs deficient in IKKα (Chuk−/−; called 'Ikka−/−' here; Fig. 1e and Supplementary Fig. 1). Therefore, overexpression of IKKβ probably caused phosphorylation of TAX1BP1 by activating endogenous IKKα. These results suggest that IKKα is essential for TAX1BP1 phosphorylation in response to stimulation with TNF or IL-1.
To determine if IKKα directly phosphorylates TAX1BP1, we did in vitro kinase assays with recombinant active IKKα and IKKβ. For substrates, we generated three different fragments of TAX1BP1 spanning the entire protein (amino acids 1–204, 205–599 and 600–747), each fused to glutathione S-transferase (GST). As a control for IKKβ phosphorylation, we used IκBα amino acids 1–54 fused to GST, a known phosphorylation substrate of IKKβ9. Of the three TAX1BP1 fragments, only the fusion protein containing amino acid 205–599 was efficiently phosphorylated by IKKα (Fig. 1f and data not shown). IKKα, in contrast to IKKβ, 'preferentially' phosphorylated TAX1BP1, whereas IKKβ 'preferentially' phosphorylated IκBα (Fig. 1f). Therefore, IKKα directly phosphorylates TAX1BP1 but IKKβ does not.
We next sought to determine if we could detect a stimulus-dependent interaction between IKKα and TAX1BP1. Endogenous TAX1BP1 interacted with wild-type IKKα, but not with the 'kinase-dead' mutant IKKα(K44M), in Ikka−/− MEFs transfected with expression vector encoding wild-type or mutant IKKα and stimulated with TNF or IL-1 (Fig. 1g). Furthermore, wild-type IKKα restored TAX1BP1 phosphorylation in reconstituted Ikka−/− MEFs stimulated with TNF or IL-1, but IKKα(K44M) did not (Fig. 1h), which suggested that the catalytic activity of IKKα is necessary for TNF- and IL-1-induced phosphorylation of TAX1BP1.
To identify the TAX1BP1 phosphorylation site(s), we undertook a combination of several approaches. First, we used a bioinformatics approach to identify consensus sites in TAX1BP1 that undergo phosphorylation by IKK. Peptide substrate array profiling has been done for all IKK family members. Quantification of that profiling has allowed both a positive matrix and negative matrix to be generated. This matrix can enable proteome-wide bioinformatic screening of potential substrates for which the local positive and negative amino acid 'preferences' surrounding the phosphorylation site are known36,37. In general, the IKKs tended to use acidic or phosphorylated amino acids both amino (N)-terminal and carboxy (C)-terminal to the phosphorylation site36,37 (Fig. 2a). Using the matrix generated before for the IKKs, we identified a total of three potential IKK phosphorylation sites in TAX1BP1 at Ser212, Ser254 and Ser593 (Fig. 2a). These potential phosphorylation sites conformed well to known IKK sites; however, because of the rate of false-positive results with this technique, we also generated a series of constructs of TAX1BP1 with deletions at the N and C termini to examine phosphorylation of TAX1BP1 by IKKα. TAX1BP1 constructs truncated at the C terminus (amino acids 1–696 and 1–650) were phosphorylated by TAX1BP1 (Fig. 2b,c). However, further deletion of amino acids from the C terminus abrogated TAX1BP1 phosphorylation, as amino acids 1–504 and 1–360 of TAX1BP1 did not undergo phosphorylation (Fig. 2b,c). These results suggest that TAX1BP1 phosphorylation occurs between amino acids 504 and 650. Thus, IKKα phosphorylates the C terminus of TAX1BP1, which indicates that Ser593 is a potential phosphorylation site.
Phosphorylation of TAX1BP1 at Ser593 and Ser624
Next we undertook an unbiased approach using mass spectrometry to identify TAX1BP1-phosphorylation site(s). We transfected 293T human embryonic kidney cells with expression vector encoding Flag-tagged TAX1BP1 and treated the cells with TNF. We immunoprecipitated TAX1BP1 with antibody to Flag (anti-Flag), separated the proteins by SDS-PAGE and stained gels with silver to visualize TAX1BP1 (Fig. 2d). We excised bands and digested them with trypsin, then enriched the resultant samples for phosphorylated peptides and analyzed them by matrix-assisted laser desorption-ionization quadrupole ion–trap time-of-flight mass spectrometry. Although we did not identify any phosphorylated TAX1BP1 peptides in untreated cells, we identified two overlapping phosphorylated peptides in TNF-treated cells (Fig. 2d). Each contained a single serine at position 593, which confirmed our hypothesis that Ser593 is indeed a TNF-inducible TAX1BP1-phosphorylation site.
Given that result, we used site-directed mutagenesis to substitute that serine residue with an alanine residue (TAX1BP1(S593A)). Unexpectedly, phosphorylation of TAX1BP1, observed as a shift in mobility by immunoblot analysis, was intact when IKKα was expressed together with TAX1BP1(S593A) (Fig. 2e), which suggested additional phosphorylation sites not detected by mass spectrometry. Therefore, we used the NetPhos 2.0 network for the prediction of phosphorylation sites to identify additional sites in the C-terminal region of TAX1BP1. Ser624, Ser641 and Ser652 all received high scores as potential phosphorylation-acceptor sites; these serine residues are all highly conserved in many species (human, mice, rat and zebrafish). To determine if these TAX1BP1 sites may also be phosphorylated, we generated single serine-to-alanine point substitutions as well as double mutants in combination with substitution of Ser593. We transfected the mutants into TAX1BP1-deficient MEFs and treated the cells with either TNF or IL-1 to trigger phosphorylation of TAX1BP1. The inducible mobility shift of TAX1BP1 was lost only whenever the serine at position 624 was substituted with alanine (TAX1BP1(S624A); Fig. 2f), which suggested that phosphorylation of Ser624 was solely responsible for the observed mobility shift of TAX1BP1. We obtained similar results with the TAX1BP1(S624A) mutant after overexpression of IKKα (Supplementary Fig. 2). Phosphorylation-induced shifts in mobility are typically observed when the phosphorylated serine or threonine is adjacent to a proline residue, and indeed Ser624, but not Ser593, had an adjacent proline that was highly conserved (Fig. 2g). Therefore, both Ser593 and Ser624 in TAX1BP1 were phosphorylated in response to overexpression of IKKα, although only Ser624 contributed to the shift in TAX1BP1 mobility.
Although Ser593 and Ser624 were both phosphorylated after cytokine stimulation, only Ser593 conformed to a consensus IKK site, which thus raised the possibility that a kinase distinct from IKKα phosphorylates TAX1BP1 Ser624. Although the amino acids surrounding TAX1BP1 Ser624 did not conform perfectly to any known consensus phosphorylation motifs, the Scansite motif scanner indicated a putative site for phosphorylation by the kinase Akt1 (RXRXXS/T, where 'X' indicates any amino acid, 'R' indicates arginine and 'S/T' indicates serine or threonine). However, phosphorylation of TAX1BP1 at Ser624 remained intact in MEFs doubly deficient in both Akt1 and Akt2 (Supplementary Fig. 3).
We next examined the interactions between the TAX1BP1-phosphorylation mutants and IKKα in TAX1BP1-deficient MEFs reconstituted with wild-type TAX1BP1, TAX1BP1(S593A), TAX1BP1(S624A) or TAX1BP1(S593A,S624A). Both wild-type TAX1BP1 and TAX1BP1(S593A) interacted with endogenous IKKα after TNF stimulation, but TAX1BP1(S624A) or TAX1BP1(S593A,S624A) did not (Supplementary Fig. 4). Thus, phosphorylation of TAX1BP1 at Ser624 seemed to stabilize the interaction between TAX1BP1 and IKKα, consistent with our findings reported above in which the 'kinase-dead' IKKα mutant did not stably interact with TAX1BP1 (Fig. 1g).
The TAX1BP1 plasmid encoding amino acids 1–747 used in our studies thus far corresponded to TAX1BP1 isoform 2 (in the Universal Protein Resource Knowledgebase–Swiss-Prot database), which lacks 42 amino acids (604–645) present in TAX1BP1 isoform 1 (amino acids 1–789) because of alternative splicing. To confirm that there were no additional phosphorylation sites in the larger isoform, we substituted Ser593 and Ser666 (which correspond to Ser624 in isoform 2) in TAX1BP1 isoform 1 (amino acids 1–789). Indeed, we observed a shift in the mobility of wild-type TAX1BP1 (amino acids 1–789) that was abolished for the TAX1BP1(S593A,S666A) double mutant (Supplementary Fig. 5). Therefore, Ser593 and Ser624 (Ser666 in isoform 1) seemed to be the only IKKα-mediated phosphorylation sites in TAX1BP1. We further confirmed that result by in vivo phosphorylation experiments after reconstituting TAX1BP1-deficient MEFs with either wild-type TAX1BP1 or the phosphorylation mutant TAX1BP1(S593A,S624A). Although wild-type TAX1BP1 was considerably phosphorylated after TNF stimulation, no phosphorylation of the TAX1BP1-phosphorylation mutant was evident (Fig. 2h). These results confirmed that Ser593 and Ser624 were the only sites on TAX1BP1 that were phosphorylated after stimulation with TNF.
To more easily monitor phosphorylation of TAX1BP1 at Ser593, we generated an antibody specific for such phosphorylation. We confirmed the specificity of the antibody by overexpressing TAX1BP1 or TAX1BP1(S593A) together with IKKα. The phosphorylation-specific antibody detected only wild-type TAX1BP1, not TAX1BP1(S593A), in the presence of IKKα (Fig. 2i). The phosphorylation-specific antibody also detected endogenous TAX1BP1 in control MEFs treated with TNF (Fig. 2j) or IL-1 (Supplementary Fig. 6), but not in Tax1bp1−/− MEFs, thus indicating its specificity. Furthermore, endogenous phosphorylated TAX1BP1 was also detected in primary mouse bone marrow–derived macrophages (BMDMs) treated with either TNF or IL-1 (Fig. 2k).
TAX1BP1 phosphorylation downregulates NF-κB signaling
TAX1BP1 is a critical negative regulator of NF-κB, acting by downregulating polyubiquitination of RIP1 and TRAF6 in the signaling pathways of the TNF receptor and IL-1 receptor, respectively29,30. To determine the functional relevance of TAX1BP1 phosphorylation, we reconstituted TAX1BP1-deficient MEFs with either wild-type TAX1BP1 or TAX1BP1(S593A,S624A). TAX1BP1 in either form was typically expressed in reconstituted Tax1bp1−/− MEFs in amounts similar to those of endogenous TAX1BP1 in wild-type MEFs (Supplementary Fig. 7). As expected, Tax1bp1−/− MEFs showed prolonged degradation of IκBα, phosphorylation of the kinase Jnk and K63-linked polyubiquitination of RIP1 and TRAF6 (Fig. 3a,b). Reconstitution of TAX1BP1-deficient MEFs with wild-type TAX1BP1 restored transient activation of NF-κB, phosphorylation of Jnk and ubiquitination of RIP1 and TRAF6 (Fig. 3a,b). However, the phosphorylation-defective TAX1BP1(S593A,S624A) mutant was impaired in the termination of signaling by NF-κB, phosphorylation of Jnk and ubiquitination of RIP1 and TRAF6 (Fig. 3a,b).
We also did functional studies of primary mouse BMDMs transfected with a single small interfering RNA (siRNA) specific for mouse Tax1bp1 to silence endogenous expression of TAX1BP1. We then reconstituted the cells with human TAX1BP1 (encoding either wild-type or phosphorylation-defective mutant TAX1BP1), which contains three mismatches with the mouse siRNA and therefore was not efficiently targeted by the siRNA (data not shown). Knockdown of TAX1BP1 triggered prolonged IκBα degradation in mouse macrophages (Fig. 3c), indicative of persistent NF-κB signaling. Similar to results obtained with MEFs, wild-type TAX1BP1 restored transient NF-κB signaling in macrophages, but the phosphorylation-defective mutant TAX1BP1 did not (Fig. 3c). Collectively, these results suggest that phosphorylation of TAX1BP1 is essential for downregulation of NF-κB signaling.
TAX1BP1 phosphorylation regulates targeting of A20 to TRAFs
Next we examined the role of TAX1BP1 phosphorylation in the recruitment of the A20 ubiquitin-editing complex to substrates. We reconstituted TAX1BP1-deficient MEFs with either wild-type TAX1BP1 or TAX1BP1(S593A,S624A), treated the cells for various times with TNF or IL-1 and monitored protein-protein interactions by immunoprecipitation (Fig. 4a,b). In TAX1BP1-deficient MEFs transfected with empty vector, there was a defect in the TNF- and IL-1-mediated recruitment of Itch, A20 and TAX1BP1 to TRAF2 and TRAF6 that was restored by transfection of wild-type TAX1BP1 (Fig. 4a,b). However, TAX1BP1(S593A,S624A) failed to interact with TRAF2 or TRAF6 and was also impaired in the recruitment of Itch and A20 to the TRAF proteins (Fig. 4a,b). A20 and TAX1BP1 act together to disrupt the interactions between TRAF2-TRAF6 and Ubc13 in signaling via IL-1 and TNF26. Indeed, TRAF2 and TRAF6 interacted persistently with Ubc13 in Tax1bp1−/− MEFs (Fig. 4a,b). Reconstitution with wild-type TAX1BP1 restored transient TRAF2-TRAF6–Ubc13 interactions, but reconstitution with phosphorylation-defective mutant TAX1BP1 did not (Fig. 4a,b). TNF signaling is known to induce the formation of a membrane-bound signaling complex (complex I) that consists of TRADD, RIP1, cIAP1-cIAP2 and TRAF2-TRAF5 and is essential for activation of NF-κB38. Thus, we examined the role of TAX1BP1 phosphorylation in TNF-dependent interactions among TRAF2, RIP1 and cIAP1. TRAF2 interacted persistently with both RIP1 and cIAP1 in TAX1BP1-deficient MEFs treated with TNF. These interactions were rendered transient after reintroduction of wild-type TAX1BP1 but not after reintroduction of phosphorylation-defective mutant TAX1BP1 (Fig. 4a). Thus, TAX1BP1 phosphorylation is essential for the recruitment of TAX1BP1, A20 and Itch to TRAF2 or TRAF6 in response to stimulation with TNF or IL-1, respectively.
TAX1BP1 phosphorylation assembles the A20 protein complex
Next we determined the relative contributions of TAX1BP1 Ser593 and Ser624 to the termination of NF-κB signaling by using single-point mutants for functional studies of Tax1bp1−/− MEFs. Wild-type TAX1BP1 abrogated the activation of NF-κB in response to TNF (Fig. 5a) or IL-1 (Fig. 5b), as measured by an NF-κB luciferase assay. However, single substitution of either Ser593 or Ser624 largely abolished the NF-κB-inhibitory effects of TAX1BP1, whereas combined substitution of both serine residues completely abolished the function of TAX1BP1 (Fig. 5a,b).
We then used immunoprecipitation to determine the role of TAX1BP1 phosphorylation in the assembly of the A20 ubiquitin-editing complex. We immunoprecipitated RNF11 from Tax1bp1−/− MEFs reconstituted with empty vector, wild-type TAX1BP1, TAX1BP1(S593A), TAX1BP1(S624A) or TAX1BP1(S593A,S624A) and treated with either TNF or IL-1. We then examined the RNF11 immunoprecipitates for immunoreactivity with anti-Itch, anti-A20 and anti-TAX1BP1. As expected, in TAX1BP1-deficient MEFs, the binding of RNF11 to Itch, A20 or TAX1BP1 was impaired after stimulation with TNF (Fig. 5c) or IL-1 (Fig. 5d), which highlighted the importance of TAX1BP1 as a scaffolding molecule for the A20 ubiquitin-editing complex. Reconstitution of TAX1BP1-deficient MEFs with wild-type TAX1BP1 restored the interactions among RNF11, Itch, A20 and TAX1BP1 (Fig. 5c,d). However, single-mutant TAX1BP1(S593A) or TAX1BP1(S624A) and double-mutant TAX1BP1(S593A,S624A) were each impaired in restoring the assembly of the A20 ubiquitin-editing complex in response to stimulation with TNF or IL-1 (Fig. 5c,d). Collectively, these data suggested that both Ser593 and Ser624 in TAX1BP1 serve essential and nonredundant roles in cytokine-dependent nucleation of the A20 ubiquitin-editing complex.
IKKα is required for assembly of the A20 protein complex
Because IKKα is essential for the phosphorylation of TAX1BP1, which in turn regulates protein-protein interactions between members of the A20 ubiquitin-editing complex, we hypothesized that IKKα-deficient MEFs would show a similar defect in formation of the A20 ubiquitin-editing complex. We stimulated wild-type, Ikka−/−, Ikbkb−/− and Ikbkg−/− MEFs with TNF and assessed the recruitment of A20 to RIP1 by immunoprecipitation (Fig. 6a). A20 interacted with RIP1 in response to TNF stimulation in wild-type, Ikbkb−/− and Ikbkg−/− MEFs but not in Ikka−/− MEFs (Fig. 6a). Furthermore, TNF-dependent interactions between TAX1BP1 and Itch were impaired in Ikka−/− MEFs (Fig. 6b). We also confirmed those results in wild-type MEFs transfected with a pool of siRNAs specific for IKKα. In agreement with the results obtained with IKKα-deficient MEFs, knockdown of IKKα abrogated the recruitment of A20 to RIP1 in response to stimulation with TNF (Fig. 6c). Furthermore, silencing of IKKα impaired the IL-1-dependent interactions among TAX1BP1, A20 and Itch (Fig. 6d). We also did IKKα-knockdown studies with primary mouse macrophages to examine TNF- and IL-1-mediated NF-κB signaling events. As expected, TNF triggered transient K63-linked polyubiquitination of RIP1 and degradation of IκBα in the presence of control siRNA (Fig. 6e). However, siRNA-mediated silencing of IKKα elicited persistent RIP1 polyubiquitination and IκBα degradation (Fig. 6e), which emphasized the importance of IKKα in terminating NF-κB signaling in primary macrophages. We obtained similar results for the IL-1 receptor pathway, as IKKα silencing triggered persistent K63-linked polyubiquitination of TRAF6 and degradation of IκBα (Fig. 6f). Thus, IKKα served an essential role in the negative feedback of NF-κB signaling by facilitating TAX1BP1-mediated assembly of the A20 ubiquitin-editing complex (Supplementary Fig. 8).
Downregulation of NF-κB signaling by phospho-mimetic TAX1BP1
Thus far our results had defined an NF-κB negative feedback loop initiated by IKKα phosphorylation of TAX1BP1 that was essential for the assembly of the A20 ubiquitin-editing complex. As TAX1BP1 is downstream of IKKα, we hypothesized that a TAX1BP1 mutant that mimics phosphorylated TAX1BP1 (phospho-mimetic TAX1BP1) would bypass the requirement of IKKα in the A20 negative feedback loop. Thus, we substituted the serine residues at positions 593 and 624 with glutamic acid (to mimic phosphorylated serine) to generate the phospho-mimetic mutant TAX1BP1(S593E,S624E). We reconstituted Tax1bp1−/− MEFs with wild-type TAX1BP1, TAX1BP1(S593A,S624A) or TAX1BP1(S593E,S624E), then transfected the cells with control siRNA with scrambled sequence or IKKα-specific siRNA and stimulated them with TNF or IL-1. As expected, the phosphorylation-defective mutant TAX1BP1(S593A,S624A) was impaired in terminating NF-κB signaling, whereas the phospho-mimetic mutant TAX1BP1(S593E,S624E) resembled wild-type TAX1BP1 in restoring transient activation of NF-κB (Fig. 7a,b). In the absence of IKKα, wild-type TAX1BP1 was unable to undergo phosphorylation and was therefore impaired in the ability to turn off NF-κB after stimulation with TNF or IL-1 (Fig. 7a,b). However, the phospho-mimetic mutant TAX1BP1(S593E,S624E) retained its ability to inhibit NF-κB in the absence of IKKα, as observed by transient degradation of IκBα (Fig. 7a,b). Therefore, the phospho-mimetic mutant bypassed the requirement for IKKα to inhibit NF-κB signaling.
HTLV-I Tax inhibits phosphorylation of TAX1BP1
TAX1BP1 was originally identified in a yeast two-hybrid screen as a protein that interacts with the HTLV-I Tax oncoprotein39,40. Studies have indicated that Tax counteracts the adaptor function of TAX1BP1 to promote persistent NF-κB signaling26,35; however, it has remained unclear if Tax affects the phosphorylation of TAX1BP1. Therefore, we transfected MEFs with plasmids encoding TAX1BP1 and Tax, then stimulated the cells with TNF or IL-1 to induce phosphorylation of TAX1BP1. As expected, TAX1BP1 was phosphorylated after stimulation with TNF or IL-1, as shown by its shift in mobility by immunoblot analysis; however, Tax completely blocked TAX1BP1 phosphorylation (Fig. 8a,b). To address the mechanism of inhibition, we examined TAX1BP1-IKKα interactions in the presence or absence of Tax. Stimulation with either TNF or IL-1 promoted interaction between TAX1BP1 and IKKα that was impaired in Tax-expressing cells (Fig. 8c). Therefore, Tax prevented the recruitment of IKKα to TAX1BP1 and the subsequent phosphorylation of TAX1BP1 in cells stimulated with TNF or IL-1. As HTLV-I infects mainly T lymphocytes in vivo, we next examined TAX1BP1 phosphorylation in a panel of HTLV-I-transformed T cell lines, as well as the Jurkat Tax Tet-on human T lymphocyte cell line that conditionally expresses Tax. Notably, TAX1BP1 phosphorylation was evident in TL-OM1 cells (a human lymphocytic adult T cell leukemia cell line that lacks Tax expression) but not in the Tax-expressing HTLV-I-transformed human T lymphocyte cell lines MT-2 and C8166 (Fig. 8d). Furthermore, treatment of Jurkat Tax Tet-on cells with the phorbol ester PMA in combination with the calcium ionophore ionomycin induced phosphorylation of TAX1BP1, but this did not occur when Tax expression was activated by treatment with doxycycline (Fig. 8d). Therefore, TAX1BP1 phosphorylation was inhibited by Tax in T cells and in HTLV-I transformed cell lines that expressed Tax despite the persistent NF-κB signaling in such cells.
A published study has indicated that siRNA-mediated silencing of human TAX1BP1 in 293T cells impairs Tax-induced NF-κB activation41, although the mechanism remains unclear. As we had found that the phospho-mimetic mutant TAX1BP1(S593E,S624E) bypassed the requirement for IKKα to terminate NF-κB signaling (Fig. 7), we next examined the effect of that mutant on Tax-mediated NF-κB activation. We reconstituted TAX1BP1-deficient MEFs with wild-type TAX1BP1 or TAX1BP1(S593E,S624E) and analyzed them by Tax-driven NF-κB luciferase assays. Indeed, activation of NF-κB by Tax was impaired in the absence of TAX1BP1 (Fig. 8e). Furthermore, Tax was unable to activate NF-κB in the presence of TAX1BP1(S593E,S624E) (Fig. 8e). This result suggested that negative feedback inhibition by TAX1BP1(S593E,S624E) may have overridden the activation of NF-κB by Tax.
Our results have demonstrated the importance of TAX1BP1 phosphorylation in the termination of TNF- and IL-1-mediated NF-κB signaling. TAX1BP1 has two C-terminal ubiquitin-binding Cys2His2 zinc finger motifs that are essential for inhibition of NF-κB30. TAX1BP1 inducibly interacts with the HECT domain–containing E3 ligase Itch via two C-terminal PPXY motifs (where 'X' is any amino acid) positioned in the center of each of the zinc-finger domains35. We hypothesize that TAX1BP1 is normally in an inactive conformation whereby the zinc-finger domains and associated PPXY motifs are inaccessible to cofactors such as Itch, RNF11 and A20 as well as polyubiquitinated substrates. However, phosphorylation of TAX1BP1 by IKKα probably triggers a conformational change in TAX1BP1 that exposes the zinc fingers and PPXY motifs to allow the recruitment of Itch, RNF11 and A20 and the concomitant recognition of substrates. The scaffolding function of TAX1BP1 is strictly dependent on IKKα-mediated phosphorylation of TAX1BP1. This mechanism of autoregulation ensures that IKK is activated only transiently by proinflammatory cytokines and raises the possibility of dysregulation of TAX1BP1 phosphorylation in inflammatory disorders and cancers in which NF-κB is commonly persistently activated.
The HTLV-I oncoprotein Tax promotes persistent NF-κB signaling through constitutive activation of IKKα and IKKβ33. Therefore, it was unexpected that Tax selectively blocked phosphorylation of TAX1BP1 by IKKα yet promotes IKKα-mediated phosphorylation of NF-κB2 in the noncanonical pathway42. Tax blocks the interactions of A20 with its substrates RIP1, Ubc13 and TRAF6 (refs. 26,35). As TAX1BP1 phosphorylation seems to be upstream of all these events, Tax inhibition of TAX1BP1 phosphorylation is probably the initial and defining event in the inactivation of A20. Also, because Tax blocked cytokine-induced interactions between TAX1BP1 and IKKα, it is plausible that Tax relocalizes TAX1BP1 to a distinct subcellular compartment inaccessible to IKKα or, alternatively, Tax may compete with IKKα for binding to TAX1BP1. Further studies are necessary to elucidate the precise role of TAX1BP1 in the activation of NF-κB by Tax.
As IKKα is a critical component of the noncanonical NF-κB pathway, it was unexpected that it also has a prominent role in the regulation of the canonical NF-κB pathway. However, it is known that extensive crosstalk exists between the canonical and noncanonical NF-κB pathways. For example, TRAF3-deficient cells have higher expression of NIK and constitutive activation of the noncanonical pathway, as well as enhanced activation of canonical NF-κB signaling43. Thus, TRAF3 negatively regulates IKK complexes in both canonical and noncanonical NF-κB pathways. A published study has demonstrated that IKKα phosphorylates NIK in negative feedback regulation of the noncanonical pathway44, which supports the idea that IKKα has important roles in terminating both the canonical and noncanonical pathways. Understanding how IKKα is regulated in the canonical NF-κB pathway will require further investigation.
Given the essential role of IKKα, but not of IKKβ, in phosphorylating TAX1BP1, it is notable that IKKα has been suggested to be a negative regulator of canonical NF-κB signaling and inflammation. Macrophages from Ikka−/− mice or knock-in mice expressing inactive IKKα show greater production of proinflammatory cytokines as a result of enhanced IKKβ kinase activation and IκBα degradation45,46. Thus, the catalytic activity of IKKα is essential for the resolution of inflammatory signaling. IKKα has been shown to phosphorylate PIAS1, the inhibitor of activated transcription factor STAT1, as a mechanism for inhibiting nuclear NF-κB and downregulating proinflammatory signaling47. Therefore, an important and nonredundant function of IKKα is to turn off inflammatory signaling pathways by phosphorylating multiple substrates that together coordinate the repression of NF-κB signaling.
Cells and reagents.
Tax1bp1+/− and Tax1bp1−/− MEFs have been described26,29; Ikka−/−, Ikbkb−/− and Ikbkg−/− MEFs were provided by M. Karin; MEFs doubly deficient in both Akt1 and Akt2 were provided by M. Birnbaum; 293T cells were from American Type Culture Collection; BMDMs were made as described26; TL-OM1 cells were provided by M. Maeda48; C8166 and MT-2 cells have been described49; and Jurkat Tax Tet-on cells were a gift from W. Greene50. Plasmids Flag-TAX1BP1 and pCMV4-Tax have been described35. The Flag-tagged TAX1BP1 deletion mutants consisting of amino acids 1–696, 1–650, 1–504, 1–360 and 546–747 were generated by PCR-assisted subcloning into p3xFlag-CMV-7.1 (Sigma). Flag-TAX1BP1 1–789 was a gift from E. Androphy51. TAX1BP1 amino acids 1–204, 205–599 and 600–747 fused to GST were generated by cloning of PCR-amplified TAX1BP1 fragments into pGEX4T3 (GE Healthcare Life Sciences). Flag-tagged point mutants TAX1BP1(S593A), TAX1BP1(S593E), TAX1BP1(S624A), TAX1BP1(S624E), TAX1BP1(S641A) and TAX1BP1(S652A) were generated with the QuikChange site-directed mutagenesis kit (Stratagene). Primer sequences are in the Supplementary Methods. Hemagglutinin-tagged IKKα, IKKα(K44M), IKKβ and IKKγ (NEMO), and IκBα amino acids 1–54 fused to GST, were provided by S.C. Sun. The siGENOME SMARTpool siRNA specific for IKKα and scrambled control siRNA were from Dharmacon–Thermo Scientific. ON-TARGETplus siRNA for mouse TAX1BP1 was from Dharmacon–Thermo Scientific. The sequence of the TAX1BP1-specific siRNA was 5′-GAACGAUGCUUCAAUAAAU-3′. Recombinant active IKKα fused to GST and histidine-tagged IKKβ were from Millipore. Recombinant TAX1BP1 amino acids 1–204, 205–599 and 600–747 fused to GST were expressed and purified from BL21 Escherichia coli competent cells with a GST Spin Trap kit (GE Healthcare Life Sciences). The following antibodies were used: anti-TAX1BP1 (ab22049; Abcam), anti-β-actin (AC15; Abcam), anti-actin (A2103; Sigma), anti-RNF11 (ab57180; Abcam), anti-cIAP1 (ab2399; Abcam), anti-UbcH5c (ab58251; Abcam), anti-Flag (M2; Sigma), anti-hemagglutinin (12CA5; Roche), anti-IκBα (C-21; Santa Cruz Biotechnology), antibody to phosphorylated IκBα (14D4; Cell Signaling), anti-TRAF2 (C-20; Santa Cruz Biotechnology), anti-TRAF6 (H-274; Santa Cruz Biotechnology), anti-IKKα (B-8; Santa Cruz Biotechnology), anti-GST (Z-5; Santa Cruz Biotechnology), anti-Jnk (56G8; Cell Signaling), antibody to phosphorylated Jnk (9251; Cell Signaling), anti-RIP1 (38; BD Biosciences Pharmingen), anti-A20 (E5-1619; BD Biosciences Pharmingen), anti-A20 (H-100; Santa Cruz Biotechnology), anti-Itch (32; BD Biosciences Pharmingen), anti-Ubc13 (clone 4E11; Invitrogen) and antibody specific for ubiquitin Lys63 (HWA4C4; Millipore). Anti-Tax35 was prepared from a Tax hybridoma (168B17-46-34) from the AIDS Research and Reference Program of the National Institute of Allergy and Infectious Diseases (US National Institutes of Health). Recombinant TNF and IL-1 were from R&D Systems; LPS was from Sigma; and calf intestinal alkaline phosphatase was from New England Biolabs.
The antibody specific for TAX1BP1 phosphorylated at Ser593 was generated by 21st Century Biochemicals by standard procedures. The phosphorylated peptide NYKELKR(pS)LENPAER (where (pS) is the phosphorylated serine residue) was used for immunization of rabbits. The phosphorylation-specific antibody was affinity purified and used for immunoblot analysis at a dilution of 1:1,000.
In vivo 32P labeling.
MEFs were washed in phosphate-free RPMI medium (Invitrogen), labeled for 1 h with 32P-orthophosphate (1 mCi/ml), treated for 30 min with TNF (10 ng/ml) and lysed in radioimmunoprecipitation buffer. Proteins were immunoprecipitated from lysates with anti-TAX1BP1, separated by SDS-PAGE, transferred to nitrocellulose membranes and visualized by autoradiography.
In vitro kinase assays.
Recombinant active IKKα fused to GST or histidine-tagged IKKβ (50 ng of either) together with TAX1BP1 amino acids 1–204, 205–599 or 600–747, or IκBα amino acids 1–54, fused to GST (1 μg of each), were used for in vitro kinase assays in a mixture containing 10 mM ATP, 1 M dithiothreitol and kinase assay buffer (20 mM HEPES, pH 7.6, 20 mM MgCl2, 20 mM β-glycerophosphate and 1 mM EDTA). Kinase reactions were incubated for 30 min at 30 °C and were terminated by the addition of 5× SDS loading buffer. Samples were boiled, separated by SDS-PAGE, transferred to nitrocellulose membranes and visualized by autoradiography.
First, 293T cells were transfected with expression vector for Flag-tagged TAX1BP1 and left untreated or treated for 30 min with TNF. Then, cells were lysed and proteins were immunoprecipitated with anti-Flag and separated by SDS-PAGE. Gels were stained with ProteoSilver according to the manufacturer's instructions (Sigma). Bands corresponding to TAX1BP1 were excised and digested with trypsin, and samples were enriched for phosphorylated tryptic peptides with NuTip TiO2 micropipet tips. Mass spectrometry and analysis were done essentially as described52 with an Axima QIT matrix-assisted laser desorption-ionization quadrupole ion–trap time-of-flight mass spectrometer (Shimadzu Instruments) at the Proteomic Mass Spectrometry Lab at the University of Massachusetts. Tryptic peptides were analyzed in positive ion mode in the mid-mass range of 700–3,000 daltons. Spectra were processed with Mascot Distiller and the Mascot search engine was used for database searches (Matrix Sciences).
Cells were transfected as described35; 293T cells were transfected with FuGENE 6 (Roche), and MEFs were transfected with FuGENE HD (Roche).
Immunoblot and immunoprecipitation.
Immunoblot analysis, immunoprecipitation and ubiquitination assays were done as described53.
We thank S. Sun (University of Texas MD Anderson) for hemagglutinin-tagged IKKα, IKKα(K44M), IKKβ and IKKγ (NEMO), and IκBα amino acids 1–54 fused to GST; W. Greene (Gladstone Institute of Virology) for Jurkat Tax Tet-on cells; M. Maeda (Kyoto University) for TL-OM1 cells; M. Birnbaum (University of Pennsylvania) for MEFs doubly deficient in both Akt1 and Akt2; M. Karin (University of California San Diego) for Ikka−/−, Ikbkb−/− and Ikbkg−/− MEFs; E. Androphy (University of Massachusetts) for Flag-TAX1BP1 amino acids 1–789; and J. Leszyk for mass spectrometry and analysis. Supported by the US National Institutes of Health (RO1GM083143 and RO1CA135362 to E.W.H.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute, National Institute of General Medical Sciences or the US National Institutes of Health.
Supplementary Figures 1–8 and Methods